Where The Rubber Meets The Road: 2010 Encoder Comparison

What about hardware co-processors? I looked at all three of the products listed in that category, including the Matrox CompressHD co-processor on both Mac and Windows. CompressHD on the Mac proved faster than Compressor, and it produced better quality H.264 files. In Windows, however, CompressHD couldn’t meet my target data rate and produced lower-quality results than Adobe Media Encoder. On both platforms, it did a great job with Blu-ray and high-bandwidth H.264 encoding, for example, for uploading to YouTube.

In general, the Elemental Accelerator (which accelerates encoding using the NVIDIA graphics processing unit, or GPU) and SpursEngine Plug-in (which uses a Toshiba encoding co-processor) proved better at fast, high-bitrate encoding for Blu-ray or high-quality uploads to YouTube than for traditional streaming encoding.

Specifically, neither co-processor could match the quality of the video produced by the Adobe Media Encoder. Also, on my eight-core HP Z800, encoding to these low bitrate targets with these products took about the same amount of time as encoding with Adobe Media Encoder without acceleration. To make a long story short, other than CompressHD on the Mac, these product are probably best utilized for high-bitrate H.264 encoding, not encoding for streaming. For this reason, I won’t discuss them further in this review.

With this as background, let’s jump into our codec-specific analysis.

Producing VP6 Files
For all the talk about H.264 and Silverlight, VP6 is still the most widely used codec on the planet. From a features perspective, when choosing an encoder for VP6 production, there are two primary considerations. First is the availability of two-pass encoding, which generally provides better quality than single-pass encoding.

Second is whether the encoding tool supports both VP6-S and VP6-E formats. Briefly, VP6-E is the highest quality VP6 format, but it is slightly harder to decode than VP6-S, which offers slightly lower-quality results. In general, you should consider VP6-S for HD encodes; otherwise use VP6-E. By way of background, all VP6 encoders originally produced the equivalent of VP6-E. Later, On2 split out VP6-S as a lower-quality, easier-to-play-back option. Accordingly, if your encoding tool doesn’t support both formats, it’s producing VP6-E.

Apple Compressor doesn’t natively produce VP6 files, so it’s not included; Microsoft’s Expression Encoder is excluded for the same reason. Otherwise, the only blip is that while Rhozet Carbon Coder can produce VP6 files natively, Rhozet’s implementation doesn’t support two-pass encoding or VP6-E/S. However, you can buy On2’s Flix Exporter for less than $200 and get both of these features.

From a quality perspective, all VP6-capable encoding tools license technology from On2, and quality is very uniform among all of the encoders. This makes encoding performance the most important consideration for many producers. That being the case, let’s have a look.

Briefly, I tested encoding speed on a Mac Pro running Snow Leopard with two 2.93GHz Quad-Core Nehalem Xeon CPUs with 18GB of 1067MHz DDR3 RAM. First, I encoded a single file; then, I encoded eight files in a single instance of the program. If the encoder could open multiple instances, which both Squeeze and Flix Pro could do, I encoded eight files in eight separate instances. Note that I didn’t include Squeeze 6 comparisons for VP6 because the initial releases of Squeeze 6 had a bug that prevented me from achieving my data rate targets.

Single-file encoding time varied significantly, with Flix Pro and Episode posting the fastest times and Squeeze 5 bringing up the rear. Interestingly, none of the encoding tools encoded files in parallel—all produced the files serially, basically wasting seven of the eight cores on the Mac Pro.